Illuminant invariance from a single reflected light

This article presents an algorithm for computing an illuminant-invariant quantity inherent in a single pixel of an imaged object color. The invariance pertains for two different assumptions about the illuminant spectrum: the photoreceptor sensitivities and the reflectance spectrum of the object. For one regime the illuminant spectrum is exponential, the photoreceptor sensitivities are equal-width Gaussians, and the reflectance is also Gaussian. For the other regime the illuminant is a Wien approximation to a blackbody radiator, the photoreceptor sensitivities are narrow band in wavelength, and the reflectance is unconstrained. The existence of two regimes for the invariant testifies to its robustness. Computing this invariant over all pixels in an image will assist object-color recognition (machine-vision color constancy) without resorting to the usual assumption that illuminant variation over a scene is gradual compared to reflectance variation over that scene.